feat(deckies): generic file drops on fleet + MazeNET deckies

Extracts the docker-exec-with-base64-stdin pattern out of canary/planter
and orchestrator/drivers/ssh into a shared decnet.decky_io package.
Both consumers now delegate; the canary planter test still proves the
contract end-to-end.

Adds POST/DELETE /api/v1/deckies/files for arbitrary file drops.
Container resolution is shared with the canary path: topology_id absent
means fleet (<name>-ssh), present routes through resolve_decky_container
which picks <name>-ssh when the topology decky exposes ssh, else the
topology base container decnet_t_<id8>_<name>.

Path validation rejects relative paths and '..' traversal at the request
model layer.  Bad base64 → 400; unknown topology → 404; decky not in
topology → 422; docker exec failure → 409.
This commit is contained in:
2026-04-28 22:43:34 -04:00
parent 3fe999d706
commit 0bc4b05c73
19 changed files with 1047 additions and 176 deletions

View File

@@ -20,11 +20,8 @@ shape but speaks bytes-via-base64 over the wire.
"""
from __future__ import annotations
import asyncio
import base64
import os
import shlex
import time
from datetime import datetime, timedelta, timezone
from secrets import token_urlsafe
from typing import Any, Iterable, Optional
@@ -34,13 +31,16 @@ from decnet.bus.factory import get_bus
from decnet.canary.base import CanaryArtifact, CanaryContext
from decnet.canary.factory import get_generator
from decnet.canary.paths import default_path_for
from decnet.decky_io import (
delete_file_from_container,
resolve_topology_container,
write_file_to_container,
)
from decnet.logging import get_logger
from decnet.web.db.repository import BaseRepository
log = get_logger("canary.planter")
_DOCKER = "docker"
_TIMEOUT = 8.0
# Container suffix — matches the orchestrator SSH driver's convention
# (``<decky_name>-ssh``). Canary placement always happens through the
# ssh container because every decky has one and it carries the most
@@ -52,77 +52,16 @@ def _container_for(decky_name: str) -> str:
return f"{decky_name}{_SSH_CONTAINER_SUFFIX}"
def resolve_topology_container(
topology_id: str, decky_name: str, services: Iterable[str],
) -> str:
"""Container name to docker-exec into for a MazeNET decky.
The ssh service container (when present) wins because it carries the
most realistic filesystem layout — same rationale as the fleet path.
Otherwise we target the base container, whose name is set by
:func:`decnet.topology.compose._container_name`.
"""
if "ssh" in set(services):
return f"{decky_name}{_SSH_CONTAINER_SUFFIX}"
return f"decnet_t_{topology_id[:8]}_{decky_name}"
def _dirname(path: str) -> str:
idx = path.rfind("/")
if idx <= 0:
return "/"
return path[:idx]
async def _run(
argv: list[str], *, stdin_bytes: Optional[bytes] = None,
) -> tuple[int, str, str]:
try:
proc = await asyncio.create_subprocess_exec(
*argv,
stdin=asyncio.subprocess.PIPE if stdin_bytes is not None else None,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
except FileNotFoundError as exc:
return 127, "", f"argv[0] not found: {exc}"
try:
stdout, stderr = await asyncio.wait_for(
proc.communicate(input=stdin_bytes), timeout=_TIMEOUT,
)
except asyncio.TimeoutError:
try:
proc.kill()
except ProcessLookupError:
pass
return 124, "", "timeout"
return (
proc.returncode if proc.returncode is not None else -1,
stdout.decode("utf-8", "replace"),
stderr.decode("utf-8", "replace"),
)
def _build_plant_command(artifact: CanaryArtifact) -> tuple[str, bytes]:
"""Compose the ``sh -c`` script + stdin payload for one artifact.
Binary safety: we base64-encode on the host and stream the result
over stdin to ``base64 -d`` inside the container, so the bytes
never touch the argv (kernel ARG_MAX would reject anything larger
than ~128KB-2MB depending on the host). Both ``base64`` (coreutils)
and ``touch -d @<unix_ts>`` are present on every Linux base image
we ship, so there's no per-distro branching.
"""
encoded = base64.b64encode(artifact.content)
mtime = int(time.time() + artifact.mtime_offset)
mode_str = oct(artifact.mode)[2:]
parts = [
f"mkdir -p {shlex.quote(_dirname(artifact.path))}",
f"base64 -d > {shlex.quote(artifact.path)}",
f"chmod {mode_str} {shlex.quote(artifact.path)}",
f"touch -d @{mtime} {shlex.quote(artifact.path)}",
]
return " && ".join(parts), encoded
# resolve_topology_container is re-exported from decky_io for back-compat
# with callers (tests, deploy hook) that imported it from this module
# before the decky_io extraction.
__all__ = [
"plant",
"revoke",
"resolve_topology_container",
"seed_baseline",
"seed_baseline_topology",
]
async def _publish(
@@ -173,14 +112,12 @@ async def plant(
await repo.update_canary_token_state(token_uuid, "failed", err)
return False, err
sh_cmd, stdin_payload = _build_plant_command(artifact)
target_container = container or _container_for(decky_name)
# ``-i`` keeps stdin attached so base64 -d inside the container can
# consume the encoded payload streamed from the host.
argv = [_DOCKER, "exec", "-i", target_container, "sh", "-c", sh_cmd]
rc, _stdout, stderr = await _run(argv, stdin_bytes=stdin_payload)
success = rc == 0
error = None if success else (stderr.strip()[:256] or f"rc={rc}")
mtime = datetime.now(timezone.utc) + timedelta(seconds=artifact.mtime_offset)
success, error = await write_file_to_container(
target_container, artifact.path, artifact.content,
mode=artifact.mode, mtime=mtime,
)
if repo is not None:
if success:
@@ -199,8 +136,8 @@ async def plant(
if not success:
log.warning(
"canary.plant failed decky=%s token=%s rc=%d stderr=%r",
decky_name, token_uuid, rc, stderr[:120],
"canary.plant failed decky=%s token=%s container=%s err=%r",
decky_name, token_uuid, target_container, error,
)
return success, error
@@ -221,12 +158,10 @@ async def revoke(
the file is gone after the call (whether we deleted it or it was
already missing); only docker / container-down errors return False.
"""
sh_cmd = f"rm -f {shlex.quote(placement_path)}"
target_container = container or _container_for(decky_name)
argv = [_DOCKER, "exec", target_container, "sh", "-c", sh_cmd]
rc, _stdout, stderr = await _run(argv)
success = rc == 0
error = None if success else (stderr.strip()[:256] or f"rc={rc}")
success, error = await delete_file_from_container(
target_container, placement_path,
)
if repo is not None:
await repo.update_canary_token_state(token_uuid, "revoked", error if not success else None)